Within-condition analysis

A condition is a defined observation segment

A condition or phase is a series of observations gathered while a stated environmental arrangement is in effect. Baseline, treatment, withdrawal, comparison, and maintenance are common labels, but the label alone proves nothing about what actually occurred.

Confirm the condition's start and end, implementer, setting, materials, dose, prompts, communication access, and procedural integrity. If an arrangement changed midphase, preserve the true date and decide whether the graph needs another condition label.

The current BACB BCBA Test Content Outline includes interpreting graphed data, single-case design features, threats to validity, procedural integrity, and data-based decisions. It identifies examination content and leaves the exact review protocol to the qualified analyst and applicable context.

Verify measurement and graph integrity

Read the operational definition, unit, denominator, observation duration, and eligible-opportunity rule. Check the x-axis sequence, actual dates, y-axis scale, missing observations, data-entry history, and condition labels. A measured zero and a missing value are different facts.

Confirm that values are comparable. Raw counts from sessions of unequal length may require a rate. Percentages based on two and twenty opportunities should show their denominators. A graph that combines changed definitions, observers, settings, or access conditions needs those changes visible.

Describe level, trend, and variability

Use the whole condition and read observations in order:

• Level: the vertical position where values generally fall.
• Trend: systematic direction and rate of change across time.
• Variability: spread and sequence around the level or trend.

A mean, median, range, split-middle line, regression line, or another quantitative aid can support the description. State the method. The raw graph remains essential because identical summaries can arise from different sequences.

The What Works Clearinghouse single-case technical documentation uses level, trend, and variability as the first stage of visual analysis. It then examines immediacy, overlap, and consistency across phases. Within-condition review prepares the comparison; it does not replace it.

Stability means predictability for the question

“Stable” sometimes means values remain within a narrow horizontal band. In a broader visual-analysis sense, a steadily increasing or decreasing series can also be predictable around a trend. Define which meaning governs the phase decision.

Perfect flatness is rarely necessary. A baseline moving strongly toward the expected intervention outcome can weaken a later contrast. High scatter can make prediction difficult. A stable adverse pattern may still require prompt clinical or safety action. Design quality, client welfare, expected latency, feasibility, and applicable requirements all affect timing.

Avoid changing a phase solely because a preferred result appears or delaying a needed action while waiting for visually perfect data. Prespecify design rules when possible and document any exception with its responsible decision-maker.

A fictional within-phase review

A fictional condition contains six observations of independent help messages during ten eligible opportunities: 3, 4, 4, 5, 5, and 6.

The mean is 4.5 and the range is 3. The first three observations average 3.67; the final three average 5.33, a difference of +1.66 after rounding. The sequence rises with little point-to-point scatter around that direction.

Calling the condition “average 4.5” would omit its upward trend. Calling it “unstable” without defining stability would also be incomplete. A clearer record says: moderate level, positive trend, limited variability around the trend, and reduced certainty that a future value will remain near the whole-phase mean.

That description supports a tentative prediction. It offers no comparison condition and no causal conclusion. Check whether opportunities, partner response, prompts, communication access, health, and setting remained comparable before using it in a phase decision.

Use a reproducible review record

The authors of Systematic Protocols for Visual Analysis found meaningful variation among published protocols. A single-case design analysis review similarly separates within-phase review from between-phase and design-level reasoning.

A review of visual-analysis reliability and training treats disagreement among observers as a recognized challenge. Record the protocol, eligible observations, calculations, independent judgments, and resolution of disagreement. Agreement can improve consistency while remaining separate from validity.

A practical sequence

1. Define the condition and verify actual implementation.
2. Lock the measure, unit, denominator, and eligible data.
3. Read the series in chronological order.
4. Describe level, trend, variability, gaps, and outliers.
5. Define stability or predictability for the specific question.
6. Check health, access, context, integrity, and concurrent events.
7. State the tentative prediction and uncertainty.
8. Record the qualified phase or review decision and its basis.